Seasonality of the flowers and ornamental plants market in São Paulo state: the case of CEAGESP-SP

Abstracts

O presente artigo teve como objetivo principal avaliar o comportamento sazonal dos volumes e preços praticados na floricultura, através do uso de periodogramas e metodologia de Box e Jenkins (1976). Para tal, viu-se a necessidade de seleção dos produtos e entreposto a serem trabalhados. Os produtos escolhidos foram: rosa, crisântemo e violeta. Quanto ao entreposto, dados referentes à década de 90 foram levantados na CEAGESP-SP (Companhia de Entrepostos e Armazéns Gerais de São Paulo). Os resultados deste trabalho apontaram períodos de pico para a movimentação e para os preços praticados dos produtos escolhidos naquele entreposto. Informações a respeito do período sazonal são de extrema importância para o sistema de comercialização de cada produto agrícola. Tal fato não é diferente com relação às flores e plantas ornamentais. Tanto produtores como consumidores podem se beneficiar a partir do conhecimento do comportamento mais sistemático que seus produtos de interesse revelam. Conhecendo os picos sazonais, o produtor pode organizar sua produção de forma a intercalar e aumentar seus picos de comercialização.

flores e plantas ornamentais; sazonalidade; periodograma e Box e Jenkins (1976)

Seasonality of the flowers and ornamental plants market in São Paulo state - the case of CEAGESP-SP¹

Roberta Wanderley da Costa Marques^I; José Vicente Caixeta Filho^II

^IAgronomic Engineer, Master in Applied Economics, Department of Economics, Management and Sociology of ESALQ-USP (E-mail: [e-mail])

^IICivil Engineer, Ph.D., Associate Professor of the Department of Economics, Management and Sociology of ESALQ-USP (E-mail: [e-mail])

ABSTRACT

This study evaluates the seasonal behavior of floriculture volumes and prices series covering most of the 1990s. The evaluation was accomplished using a periodogram and Box and Jenkins (1976) methodology. Three floral products were chosen for evaluation: rose, chrysanthemum, and violet. Data for the 1990’s were collected from CEAGESP-SP – Companhia de Entrepostos e Armazém Gerais de São Paulo, a floral products trade center. Study results show that traded volumes and prices peaked in particular seasons. Information regarding this seasonal period is of extreme importance to the flower and ornamental plant trading systems and allows producers to manage production to increase output at trading peaks.

Key words: Flowers and ornamental plants; seasonality; periodogram; Box and Jenkins.

1. Introduction

Floriculture, in the broad sense, involves the cultivation of flowers and ornamental plants for various purposes, which range from the culture of flowers for trading to the production of arboreal seedlings.

According to IBRAFLOR – Brazilian Institute of Floriculture (2001), the total area devoted to the production of flowers and ornamental plants in Brazil reached 4,850 ha in 1999, with the states of São Paulo, Minas Gerais and Rio de Janeiro together accounting for 75% of national production.

Kras (1999) determined that 90% of the production and consumption of flowers and ornamental plants takes place within 500 km, as distribution and transportation costs and the product’s highly perishable nature limit trading distances. The trading and distribution of flowers and ornamental plants generally occurs in large trade centers.

Castro (1998) found three types of floriculture trade centers: those where only products from producers are bought and sold, the case of Veiling Holambra and Mercaflor-SC (Mercado do Profissional de Floricultura e do Paisagismo); marketplaces where products from both producers and wholesalers are traded, such as CEAGESP-SP (Companhia de Entrepostos e Armazéns Gerais de São Paulo) and CEASA-Campinas (Central de Abastecimento S/A); and marketplaces where only wholesalers trade, such as CADEG - Rio de Janeiro (Companhia de Abastecimento do Estado da Guanabara) and CEASA - Porto Alegre.

The main floriculture trading centers in the state of São Paulo are Veiling Holambra, CEAGESP-SP, and CEASA-Campinas. Research carried out by IBRAFLOR in 1995, published by Arruda et al. (1996), determined that CEASA-Campinas and CEAGESP-SP account for 53% of the flower trading in Brazil.

Referring to the flower market of CEAGESP-SP in the city of São Paulo, Arruda et al. (1996) point out that by concentrating the production of several regions, the trade center facilitates flower buyers’ activities. However, the author found that flower trading in these large centers is negatively impacted by the lack of a flower classification system and long trading periods, which damage the products’ appearance

Many authors have noticed that there are seasonal changes in the consumption of flowers and ornamental plants in Brazil. According to Almeida and Aki (1995), flower production is concentrated during times of higher demand, such as Mother’s Day, All Soul’s Day, and Christmas. Claro (1998) remarks that other days of high flower consumption have recently appeared, such as International Women’s Day, Valentine’s Day, Grandparents’ Day, and Father’s Day.

According to Groot (1999), worldwide flower sales were approximately US$ 12.5 billion in 1985, rising to US$ 25 billion in 1990 and US$ 31 billion in 1995. Barletta (1995) found that annual Brazilian per capita expenditures on flowers was US$ 4.00 in 1994 while Castro (1998) found this figure to be US$ 7.00 in 1998.

Groot (1999) and Kras (1995), among others, believe that worldwide flower demand will increase; though, production will increase even faster than consumption, initiating increased price competition. However, Salunkhe et al. (1990) and others believe that improved flower marketing will cause consumer flower and ornamental plant demand to always outpace production.

According to a study carried out by Gatti (1988), a dependable supply of floriculture products throughout the year would be of benefit to both consumers and producers. For consumers, regular product supply and the elimination of product shortages would result in more accessible prices, less price variation, and predictable product availability. For producers, regular supply would result in more stable prices, more predictable incomes, increased production efficiencies, and enhanced consumer demand.

Reaching the goal of price stability in the agricultural product markets is made all the more difficult by the sectors inherent characteristics. As Santiago et al. (1996) remark, the agricultural market characterizes itself by showing a higher degree of price elasticity when compared to the industrial goods market. This leads to a wide range of price variation.

Flower prices and supply are greatly affected by Brazilian distribution sector inefficiencies, which have not improved relations between Brazilians consumers and the floriculture market. These inefficiencies and the fact that consumption is concentrated into specific periods exacerbate the difficulties encountered when trying to guarantee consistent year-round flower and ornamental plant supply. As was found by Matsunga (1995), the seasonal nature of flower production itself makes consistent flower production difficult. Of course, technical solutions, such as cultivation in greenhouses, would make continuous production more feasible.

Pino et al. (1994) point out that seasonality is determined by two major factors: the year’s climatic seasons and cultural factors that affect demand for certain products on specific dates and over specific periods in the calendar year. According to the same authors, the calendar itself causes seasonality through the change in flow variables, such as variation in the number of days in the month. Granger et al. (1978) remark decisions made by institutions and individuals are affected by calendar dates, such the timing of school vacations, the end of a fiscal year considerations, and the expectation of increased sales and demand on specific calendar dates.

If the effects and timing of factors that influence seasonality are well-known, more effective market forecasting will be possible. Knowledge of seasonal price variations can help consumers determine the best times to buy and permit producers to efficiently schedule production. Thus, it is important that empiric studies are conducted that provide accurate information to aid in the forecast of seasonal agricultural market fluctuations, in this case seasonal fluctuations in the Brazilian flowers and ornamental plants market.

This study is intended to identify the seasonal behavior of prices and traded volumes in the floriculture market. To accomplish this, rose, chrysanthemum, and violet price and volume data were collected for the 1990s from CEAGESP-SP.

2. Material and method

2.1. Data specifications

This study used monthly data on floriculture product traded volumes and prices from mid-1992 to mid-2001. The data were gathered from local researchers and from spreadsheets published in the Monthly Report of the CEAGESP-SP floral products trade center. Due to the asymmetry of information, the irregularity of data collection on the part of the trade center, and the occasional insertion of new products into the market, the periods in which information was collected for a specific product are different: the period of analysis for roses was from April 1992 to October 1999; the period for chrysanthemums was from May 1993 to December 1999; and the period for violets was from October 1993 to October 1999.

2.2. Approaches used to treat seasonality

According to Box and Jenkins (1976), spectral analysis can be used to analyze a time series. This type of analysis decomposes the series of data into sinusoidal components with random, uncorrelated coefficients. Concurrent with decomposition into sinusoidals there is a corresponding decomposition of the auto co-variance. Thus, the spectral decomposition of a stationary process is analog to the Fourier representation of deterministic functions. Fourier analysis decomposes the series into a sum of sine waves and of co-sine waves of different amplitudes and wavelengths.

The periodogram is a tool of extreme importance in this study. One of its uses is to detect the seasonal period of a time series. Box and Jenkins (1976) state that the periodogram is an appropriate tool to analyze composed time series that involve sine and co-sine waves with fixed frequencies buried in noise, as defined in the equation below:

where:

I is the contribution of frequency l _k to the adding of squares associated to the sine and co-sine coefficients;

I_k is the Fourier frequency;

dc is the transformed co-sine of Fourier

ds is the transformed sine of Fourier

When I (I l_k ) are plotted against I l_k, periodograms are formed, or else, spectrum estimators.

The Fisher periodicity test (1929) is used to check if the period found is significant, as cited by Wei (1989). This test compares the peak values and total variability of the series to detect significance.

Fisher (1929), cited by Wei (1989), developed a test based on the g statistic. According to null hypothesis Ho that periodicity does not exist, the g value is determined – shown in equation (2) – and rejects Ho if the calculated g value is higher than the g tabled (see: Wei, 1989, p. 262).

being:

N = number of observations

I = periodogram

Thus, a significant value of I(l _j) leads to the rejection of Ho and proves that the periodic component is significant; otherwise, this component would not exist. Rejecting Ho means that the series shows a known periodicity , where l* is the frequency corresponding to maximum I(l _j).

Morettin and Toloi (1987) point out that for the case of a non-stationary series, the non-stationarity characteristic of homogeneous series should be discharged before proceeding with this analysis.

2.3. Box and Jenkins methodology

The methodology and models used in this study were developed by Box et al. (1994) to treat a time series. According to several authors, among them Harvey (1993), Morettin and Toloi (1987), Montello (1970), Pino et al. (1994), a time series can be defined as a group of observations ordered in time.

The models to be specified in this study can be applied to series that show stationary and non-stationary characteristics. Autoregressive (AR), moving average (MA), or mixed autoregressive/moving average (ARMA) models can be applied to stationary series, whereas, integrated mixed autoregressive/moving average (ARIMA) models and seasonal integrated mixed autoregressive/moving average (SARIMA) models can be applied to non-stationary series. The ARMA model has order p for the autoregressive parameters and q for the moving averages parameters, while the ARIMA model has order p, d and q, d as the integration order of the model. It is possible that the series under analysis shows several seasonal components, and to express this multiple seasonality, the model must be elaborated in a way to include various stages of difference and several seasonal moving average operators and seasonal autoregressive operators.

For each of these models, the order can vary. Following the parsimony principle of Box and Jenkins (1976), the model that represents the fewest possible parameters is chosen

According to Box and Jenkins (1976), Morettin and Toloi (1987) and other authors, SARIMA models can be represented as follows:

with:

and

where:

= distinct variable Zt minus its own average ( m );

f (B) = autoregressive operator of order p;

q (B) = moving average operator of order q;

f (B) = autoregressive seasonal operator;

Q (B) = seasonal operator of moving average;

Q and P are the parameter orders of moving averages and autoregressive parameters respectively;

s is the seasonal period.

Cunha and Margarido (1999) point out that , where D^d is the difference operator, that is, , where z_t is the variable in level and B is the operator in retard ().

According to Box and Jenkins (1976), the resulting multiplication process can then be de designated as of order ( p, d, q) x (P, D, Q)s.

Following the specification of models to be used in the analysis, the series generator process is identified: is the process autoregressive or of moving averages. This means that the p, d, and q values of model ARIMA (p, d, q) are determined in this phase. If the series has already been filtered, the model is then identified as integrated autoregressive, integrated moving averages (ARIMA), or seasonal integrated moving averages (SARIMA).

Parameters of the newly identified model are then estimated. According to Gujarati (2000), this phase is today carried out using several statistic packages developed for computers.

Finally, through a study of residues, the model is analyzed to determine if it meets the desired criteria. Box and Jenkins (1976) and Morettin and Toloi (1987) have suggested a number of tests to verify that the estimated model is properly adjusted to the data, such as the residual auto-correlation test and the Box-Pierce test.

The analysis presented in this study was conducted using SAS software (1999) in accordance to specific methodology found in SAS (1993) and SAS (1996).

To summarize, volume and price data from CEAGESP-SP for each of the three chosen products were analyzed in the following order.²

a) determining the simple regression between the average and the standard series deviation to identify the kind of transformation (generally, logarithmic) which would be appropriate for the homogenization of residue variances;

b) from a transformed series, the auto-correlation function (FAC) was observed and, noticing the existence of a stochastic trend of an eventual seasonal behavior, the test of unique root was carried to check the characteristics of a stationary series;

c) if the series carried a unit root, an order difference "1" was made to eliminate the existing stochastic trend;

d) after eliminating the eventual stochastic trend, the seasonal period was determined from the peridogram (generally with a difference) and validated with the Fisher periodicity test, covered in Section 2.2;

e) variance, standard deviation, results from the auto correlation residues, and the parameters estimated according to the adjustment of equations covered in section 2.3 were then verified and evaluated using the Box and Jenkins test, the results from the auto and partial auto correlation function of the series (FAC and FACP), and Akaike’s and Schwarz’s criteria.

3. Results and discussion

Charts 1 and 2 show the volume and price series for roses, chrysanthemums, and violets at CEAGESP-SP. Over the period studied, it is observed that roses are the most traded product and that a box of violets is the most expensive.

Chart 1
- Click to enlarge

Chart 2
- Click to enlarge

Between 1998 and 1999, the common chrysanthemum was the main flower income source for traders at CEAGESP-SP. A trend test shows that the volume of chrysanthemums traded between October 1993 and October 1999 tended to increase; although, trading significantly decreased between 1995 and 1996. It is also observed that in 1994 the market supply of chrysanthemum diminished, especially in November. A general chrysanthemums price increase is noted between 1993 and 1994, followed by a major price decrease between 1995 and 1996 and a period of relative price stability from 1998 to 1999.

Cut roses accounted for the second highest flower trading income between 1998 and 1999 at CEAGESP-SP. Between April 1992 and October 1999, there was a large but decreasing volume of roses traded at the marketplace. The trading decrease is due, in part, to competition from other cut flower species and a growing consumer preference for potted flowering plants, which last longer and are more practical for use in ornamental plantings. Over the same period, there was an increase in the traded volume of violets, lilies, anthuriums, and gerberas with prices peaking in June and July of each year: probable seasonal behavior. The monthly rose price series shows a slight price increase over the study period.

Violets are representative of a great many of the flowers available in the CEAGESP-SP flower market. Violets are traded in small pots, which are sold in 6 kg boxes (containing approximately 15 pots). From 1994 to 2000, around 400 thousand boxes of violets were traded per year at CEAGESP-SP. Over the period from October 1993 to October 1999, violet trading tended to increase, although there was a major trading decrease between 1995 and 1996 and a recurrent decline in trading was noted each November. The violet price series shows a sharp fall in 1996.

Important volumes of roses, chrysanthemums, and violets are traded in the months of May, October, and December. Chart 3 presents the price and volume series peak periods at CEAGESP-SP. The traded volume of roses and chrysanthemums has a peak period of 6 months. Violet demand was found to be more consistent throughout the year, with a peak period of 3 months. The peak price period for chrysanthemums and violets is 6 months. The peak rose price period was found to be 12 months.

Chart 3
- Click to enlarge

Chart 3 also summarizes the equations obtained according to Box and Jenkins methodology. For example, the highest rose price peak at CEAGESP-SP was every 12 months, and the estimated model’s equation shows that

- the price series for roses at this trade center at time t is influenced 27.09% by the price at t – 1 month, the price in previous month;

- the seasonal self-regressive parameter shows that rose prices in period t are influenced by rose prices at t –12 months, that is, rose prices are influenced by the prices in the same month of the previous year;

- the moving average parameter of order 12 shows that rose prices at time t tend, on average, to adjust themselves to around 83.97% of the rose prices at t –12 months; therefore, it is the model’s measure of adjustment.

For all the series, flower traded volumes and prices over a period depend not only on seasonal variance, as presented in the previous section, but also on the prices and traded volumes one month and one year prior to the period’s beginning.

4. Conclusions

An understanding of the seasonal flower and ornamental plant price and trading volume peaks is essential to the diversified floriculture product producers’ strategic decision making process. This knowledge permits the producer to synchronize production to peak periods, thereby taking advantage of and possibly extending them, and may well lead to increased business’ profits and a more predictable, consistent cash flow. Higher and more regular profits mean security for the producer and generate other benefits, such labor force stability, better risk management and cost control, and the ability to plan and implement production technology improvements. If the peak periods in various countries are known, flower exportation may also become a more viable proposition.

The study results present evidence of a clear seasonal pattern in flower and ornamental plant prices and traded volumes series at CEAGESP-Sào Paulo, a large floriculture products trading center.

The volume of roses traded at CEAGESP-SP had seasonal period of 6, meaning that the most frequent volume peaks occur every 6 months. Peaks in the volume of roses traded were also observed to be more intense in the months of October, May, and December, probably due to All Soul’s Day, Mother’s Day, Christmas, and New Year’s Day. The rose price series period was found to be 12; however, this peak was not too significant, possibly reflecting this series’ lack of periodic behavior. Rose prices showed their highest peak in July, most probably due to supply constraints.

The chrysanthemum traded volume and price series periods were both 6, indicating peaks every six months. The volume of chrysanthemums traded peaked in the months of May, October, and December, again probably due to All Soul’s Day, Mother’s Day, Christmas, and New Year’s Day. Chrysanthemum prices showed increases in December, due to the high demand, and February, due to low supply.

Violet trading peaked most often every three months: a volume series peak period of 3. Consumers buy these flowers both for personal pleasure and to give as presents, which may explain why violet trading is more regular over the year than either rose or chrysanthemum trading. The violet prices series period of 6 was significant. There were price and trading peaks in the months of March, May, October, and December. The month of March appears as important new date in the violet price and volume series, probably due to International Women’s Day and the end of school vacation.

A few difficulties were encountered while conducting this study, mainly in the area of data collection. It was observed that the trade center’s data collecting activities are irregular and unsystematic.

Future effort should be made to create floriculture data series that involve other flowers and ornamental plants, studies over longer periods, or include intervention models focusing on the effect of economic shocks, plant disease, change in sector structure, or change in consumer habits.

References

• ALMEIDA, F. R. de F.; AKI, A. Y. Grande crescimento no mercado de flores. Agroanalysis, v.15, n.9, p.8-11, 1995.
• ARRUDA, S. T.; OLIVETTE, M. P. A.; CASTRO, C. E. F. Diagnóstico da floricultura do Estado de São Paulo. Revista Brasileira de Horticultura Ornamental, v.2, n.2, p.1-18, 1996.
• BARLETTA, A. Stimulating demand for flowers in Brazil. Flora Culture International, p.38-39, Sep. 1995.
• BOLETIM MENSAL GEAGESP, São Paulo: CEAGESP, 1992-2001.
• BOX, G. E. P.; JENKINS, G. M. Time series analysis: forecasting and control. San Francisco: Holden-day, 1976. 375 p.
• BOX, G. E. P.; JENKINS, G. M.; REINSEL, G. Time series analysis: forecasting and control. New Jersey: Prentice Hall/Englewoods Cliffs, 1994. 598 p.
• CASTRO, C. E. F. Cadeia produtiva de flores e plantas ornamentais. Revista Brasileira de Horticultura Ornamental, v.4, n.˝, p.1-46, 1998.
• CLARO, D. P. Análise do complexo agroindustrial das flores no Brasil Lavras, MG: UFLA, 1998. 103 f. Dissertação (Mestrado) Universidade Federal de Lavras, Lavras.
• CUNHA, M. S. da; MARGARIDO, M. A. Avaliação dos impactos dos planos de estabilização pós-1986 sobre o índice geral de preços (IGP): uma aplicação da metodologia Box & Jenkins. Agricultura em São Paulo, v.46, n.2, p.1-18, 1999.
• GATTI, E. U. A evolução recente do setor de flores e plantas ornamentais no Brasil. Agricultura em São Paulo, v.35, p.123-147, 1988.
• GRANGER, C. W. J. Seasonality: causation, interpretation and implications. In: ZELLNER, A. (Ed.). Seasonal analysis of economic time series Washington: Department of Commerce, Bureau of the Census, 1978. p.33-46. (Economic Research report, 1).
• GROOT, N. S. P. de. Floriculture worldwide trade and consumption patterns. Acta. Horticulturae, n.495, p.101-121, Sep. 1999. (Apresentado ao World Conference on Horticultural Research, Rome, 1998).
• GUJARATI, D. Econometria básica 3. ed. São Paulo: Makron Books, 2000. 846 p.
• HARVEY, A. C. Time series models Cambridge: The MIT Press, 1993. 308 p.
• IBRAFLOR. Brasil: mostra tua flora. Informativo, v. 7, n.23, p.4, mar. 2001.
• KRAS, J. Marketing of cut flowers in the future. Acta Horticulturae, n.482, p.401-405, Mar. 1999. (Apresentado ao International Symposium of Cut Flowers in the Tropics, Bogota, 1998).
• MARQUES, R. W. C. Avaliação da sazonalidade do mercado de flores e plantas ornamentais no Estado de São Paulo Piracicaba, SP: ESALQ/USP, 2002. 114 f. Dissertação (Mestrado) Escola Superior de Agricultura "Luiz de Queiroz", Universidade de São Paulo, Piracicaba.
• MATSUNAGA, M. Potencial da floricultura brasileira. Agroanalysis, v.15, n.9, p.56, Set. 1995.
• MONTELLO, J. Estatística para economistas Rio de Janeiro: APEC, 1970. 333 p.
• MORETTIN, P. A.; TOLOI, C. M. de C. Previsão de séries temporais São Paulo: Atual, 1987. 438 p.
• PINO, F. A.; FRANCISCO, V. L. F. dos S.; CÉZAR, S.A.G. et al. Sazonalidade em séries temporais econômicas: um levantamento sobre o estado da arte. Agricultura em São Paulo, v.41, n.3, p.103-133, 1994.
• SALUNKHE, D. K.; BHAT, N. R.; DESAI, B. B. Postharvest biotechnology of flowers and ornamental plants Germany: Springer-Verlag, 1990. 183 p.
• SANTIAGO, M. M. D.; CAMARGO, M. de L. B.; MARGARIDO, M. A. Detecção e análise de outliers em séries temporais de índices de preços agrícolas no Estado de São Paulo. Agricultura em São Paulo, v.43, n.2, p.89-115, 1996.
• SAS INSTITUTE: Release 8 Cary, 1999. (software).
• SAS INSTITUTE. SAS/ETS user’s guide, version 6. 2. ed. Cary, NC: [s. n.], 1993. 1022 p.
• SAS INSTITUTE. SAS/ETS Software: changes and enhancements for release 6.12. 2. ed. Cary, NC: [s. n.], 1996. 112 p.
• WEI, W. W. S. Time series analysis Redwood City: Addison-Weslwy, 1989. 478 p.

1

Article based on the Master's degree dissertation of the first author, developed under orientation of the second author.

2

Further details on tables, tests and timetable analyzed in this sequence can be found in MARQUES (2002).

Publication Dates

• Publication in this collection
  13 Dec 2004
• Date of issue
  Dec 2002